TY - JOUR T1 - Attention selectively reshapes the geometry of distributed semantic representation JF - bioRxiv DO - 10.1101/045252 SP - 045252 AU - Samuel A. Nastase AU - Andrew C. Connolly AU - Nikolaas N. Oosterhof AU - Yaroslav O. Halchenko AU - J. Swaroop Guntupalli AU - Matteo Visconti di Oleggio Castello AU - Jason Gors AU - M. Ida Gobbini AU - James V. Haxby Y1 - 2016/01/01 UR - http://biorxiv.org/content/early/2016/03/23/045252.abstract N2 - Humans prioritize different semantic qualities of a complex stimulus depending on their behavioral goals. These semantic features are encoded in distributed neural populations, yet it is unclear how attention might operate across these distributed representations. To address this, we presented participants with naturalistic video clips of animals in their natural environments while they attended to either behavior or taxonomy. We used models of representational geometry to investigate how attentional allocation affects the distributed neural representation of animal behavior and taxonomy. Attending to animal behavior transiently increased the discriminability of distributed population codes for observed actions in anterior intraparietal, pericentral, and ventral temporal cortices, while collapsing task-irrelevant taxonomic information. Attending to animal taxonomy while viewing the same stimuli increased the discriminability of distributed animal category representations in ventral temporal cortex and collapsed behavioral information. For both tasks, attention selectively enhanced the categoricity of response patterns along behaviorally relevant dimensions. These findings suggest that behavioral goals alter how the brain extracts semantic features from the visual world. Attention effectively disentangles population responses for downstream read-out by sculpting representational geometry in late-stage perceptual areas.Significance Humans can extract different kinds of high-level information from the visual world depending on their behavioral goals. Here, we use naturalistic stimuli and simple models of neural representation to investigate whether attention affects how the brain encodes semantic information. When paying attention to the behavior of an animal in its natural environment, for example, the neural representation of the observed action becomes more distinct, while irrelevant information about taxonomy is collapsed. Attending to taxonomy, on the other hand, has the inverse effect. These attentional effects occur primarily in late-stage sensorimotor areas ratherthan in early sensory areas. Overall, our behavioral goals dynamically alter how the brain processes the semantic qualities of a stimulus to better encode important information. ER -